%0 Journal Article
%T 基于Kinect的母猪呼吸频率测定算法
%A 陶源栋
%A 沈明霞
%A 刘龙申
%A 陆明洲
%A 许佩全
%A 施宏
%J 南京农业大学学报
%D 2017
%R 10.7685/jnau.201701032
%X [目的]呼吸频率是评价母猪健康的重要指标。针对目前畜牧业对母猪呼吸频率无接触测定的需求，提出一种基于Kinect的无接触测量算法。[方法]使用Kinect采集梅山母猪侧卧时的深度图像。通过动态区域检测算法确定腹部区域，采用DBSCAN密度聚类算法得到腹部区域质心进行局部深度图像提取。最后基于腹部区域局部深度图像的深度均值，使用离散傅里叶变换得到深度-时间拟合曲线，通过极值点确定1次呼吸过程的周期进而求出监测时间段内母猪的呼吸频次，最后转化成母猪的呼吸频率（RF）。[结果]将检测结果与人工计数结果做比较，呼吸频率正确率为85.3%，该呼吸频率测定算法是有效的。[结论]提出了一种呼吸频率测定方法，该方法实现无接触测量，准确率高，为定位饲养条件下的梅山母猪呼吸频率的测定提供了理论支撑。</br>[Objectives]Respiratory frequency is an important health condition indicator in sows. According to the breeding industry,there is a need for non-contact methods of determining sow’s respiratory rate. This study introduces a novel non-contact method to estimate respiration frequency algorithm based on Kinect.[Methods]Kinect was used to collect Meishan sow abdomen depth images at the time of side lying. Dynamic region detection algorithm determines the abdomen region,then the centroid of the abdomen was obtained by DBSCAN density clustering algorithm to extract the local depth image. Finally,after obtention,the Depth-time fitting curve by the discrete Fourier transform based on average depths of depths image of abdomen region was used to calculate the respiratory frequency of sow during the monitoring period through cycle calculated by the distance of extreme points.[Results]Compared to the manual counting,the correct rate of the respiratory rate calculated by proposed algorithm was 85.3%. Thus,the respiratory rate measurement algorithm worked efficiently.[Conclusions]A highly accurate respiratory frequency measurement method that can provide academic support for Meishan sow during feeding condition is proposed
%K Kinect
%K 母猪
%K 呼吸频率
%K 动态区域检测算法
%K DBSCAN密度聚类算法
%K 离散傅里叶变换&lt
%K /br&gt
%K Kinect
%K sow
%K respiratory frequency
%K dynamic region detection algorithm
%K DBSCAN density clustering algorithm
%K discrete Fourier transform
%U http://nauxb.njau.edu.cn/oa/darticle.aspx?type=view&id=201705022